Browse the corpus

Since the mid-1980s, concern has grown that invasive group A streptococcal infections (IGASI) have been increasing in incidence and severity (1–3). In particular, the emergence of streptococcal toxic shock syndrome (STSS) during the 1980s is frequently cited as an example of increasing severity (4). Person-to-person transmission of Streptococcus pyogenes (the causative agent for IGASI) primarily occurs through respiratory droplets, although it may also spread through body secretions from an infected patient (5,6). Additionally, M serotypes of S. pyogenes that cause severe disease in a patient are more likely to cause severe disease in subsequent patients (6). These serotypes include 3 (M1, M3, and M18) that are strongly associated with pathogenicity (7). Nonetheless, some evidence indicates that persons in whom IGASI from the same strain of S. pyogenes develops may have different clinical manifestations of this disease (8,9). Other risk factors for IGASI include patient’s age and underlying medical conditions (e.g., varicella). However, what factors may be associated with different clinical manifestations of IGASI is unclear (10–22).

om the same strain of S. pyogenes develops may have different clinical manifestations of this disease (8,9). Other risk factors for IGASI include patient’s age and underlying medical conditions (e.g., varicella). However, what factors may be associated with different clinical manifestations of IGASI is unclear (10–22). Some studies have examined the role of age, varicella, and chronic conditions such as diabetes mellitus and alcoholism as predictors for necrotizing fasciitis, soft-tissue infections, and STSS (21–24), yet little is known regarding other IGASI determinants. In this study, we describe the status of both IGASI and their clinical manifestations on the island of Montreal. We also identify predictors for clinical manifestations and death due to IGASI, which could explain temporal fluctuations in the incidence and severity of this disease. Methods Surveillance of IGASI Data used in our study were collected during passive surveillance of IGASI among all residents of the island of Montreal (population = 1.8 million: 21,529 births per year from 1996 to 1999 [25]). Cases that occurred and were reported from January 1, 1995 (the year IGASI became a notifiable disease in the province of Quebec), through February 28, 2002, were included in our study.

ance of IGASI among all residents of the island of Montreal (population = 1.8 million: 21,529 births per year from 1996 to 1999 [25]). Cases that occurred and were reported from January 1, 1995 (the year IGASI became a notifiable disease in the province of Quebec), through February 28, 2002, were included in our study. Once the public health department had been notified of a potential case, usually by a hospital laboratory, a 6-part questionnaire was completed by using information from the physician or infection control nurse of the health center where the case-patient was identified or treated. Questions included the patient’s demographic information, general medical information, laboratory results, diagnostic criteria, and medical history before the IGASI. With this information, all IGASI were classified into 1 of 3 groups: confirmed cases (S. pyogenes isolated from a normally sterile site), clinical cases (S. pyogenes isolated from a nonsterile site and toxic shock not attributable to any other cause), or noncases. Data on confirmed and clinical cases were entered into the regional notifiable infections computer database. This database was used for our study.

ogenes isolated from a normally sterile site), clinical cases (S. pyogenes isolated from a nonsterile site and toxic shock not attributable to any other cause), or noncases. Data on confirmed and clinical cases were entered into the regional notifiable infections computer database. This database was used for our study. Laboratory Assessment of IGASI Isolates Initial laboratory confirmation of S. pyogenes was made by using standard methods (26). Isolates were then collected and sent to the Canadian National Center for Streptococcus in Edmonton for further testing of the opacity factor, as well as M, T, and R surface proteins. The methods have been described in detail elsewhere (27). Briefly, antiopacity factor (AOF) typing was performed on any positive opacity factor sample. Although AOF testing does not possess the same type specificity as M typing, it is frequently used because of difficulties in producing antisera for certain M serotypes. Its use has been validated for most strains identified in industrialized countries. However, since 2000, the national center has supplemented AOF testing with emm gene sequencing for some nontypeable M serotype samples. Data on these results were not available for this study. Classification of Outcomes For our study, we looked at 5 dichotomous outcomes: STSS, soft-tissue infections, bacteremia, pneumonia, and death attributable to IGASI. All were invasive and defined in accordance with the classification of group A streptococcal infections (28).

Laboratory Assessment of IGASI Isolates Initial laboratory confirmation of S. pyogenes was made by using standard methods (26). Isolates were then collected and sent to the Canadian National Center for Streptococcus in Edmonton for further testing of the opacity factor, as well as M, T, and R surface proteins. The methods have been described in detail elsewhere (27). Briefly, antiopacity factor (AOF) typing was performed on any positive opacity factor sample. Although AOF testing does not possess the same type specificity as M typing, it is frequently used because of difficulties in producing antisera for certain M serotypes. Its use has been validated for most strains identified in industrialized countries. However, since 2000, the national center has supplemented AOF testing with emm gene sequencing for some nontypeable M serotype samples. Data on these results were not available for this study. Classification of Outcomes For our study, we looked at 5 dichotomous outcomes: STSS, soft-tissue infections, bacteremia, pneumonia, and death attributable to IGASI. All were invasive and defined in accordance with the classification of group A streptococcal infections (28). STSS was defined according to the 1993 Working Group on Severe Streptococcal Infections consensus definition for a probable or confirmed case (28). Soft-tissue outcomes included fasciitis, myositis, cellulitis, or erysipelas. Bacteremia was characterized by a positive hemoculture, without any source of infection. Pneumonia attributable to IGASI was based on a clinical diagnosis made by the treating physician and could include STSS with respiratory distress.

(28). Soft-tissue outcomes included fasciitis, myositis, cellulitis, or erysipelas. Bacteremia was characterized by a positive hemoculture, without any source of infection. Pneumonia attributable to IGASI was based on a clinical diagnosis made by the treating physician and could include STSS with respiratory distress. Classification of Independent Variables Age, calendar month, and year in which the IGASI case occurred were included in our study as continuous variables. Gender (male or female); underlying medical conditions (drug use, alcohol abuse, varicella, prior trauma or wound, cancer, and immunosuppression); type of living environment (hospital, daycare or preschool, school, work, other, and not available); as well as M, T, and R surface protein serotypes (presence or absence of a specific serotype) were all included as dichotomous variables. For those serotypes with identical strength of association with a given outcome, a single new dichotomous variable was created to represent the presence of one or the other (e.g., presence of either serotype M12 or M28 versus absence of both serotypes). Finally, since predominant site of infection (bacteremia, fasciitis, cellulitis or erysipelas, myositis, peritonitis, respiratory manifestations, septic arthritis, and other) was partially used in distinguishing between bacteremia, pneumonia, and soft-tissue infections, this variable was only considered a covariate of interest in models with STSS and death as their outcomes.

fasciitis, cellulitis or erysipelas, myositis, peritonitis, respiratory manifestations, septic arthritis, and other) was partially used in distinguishing between bacteremia, pneumonia, and soft-tissue infections, this variable was only considered a covariate of interest in models with STSS and death as their outcomes. Statistical Analysis The incidence (per 100,000), death rate attributable to IGASI, and proportion of IGASI cases due to a specific clinical manifestation were estimated by using data collected from 1995 through 2001. Incidence and proportion estimates were not calculated for 2002, given that only 2 months of data were available. Projected annual population estimates for Montreal were used when calculating the reported annual incidence of IGASI (25). Incidence and proportion of IGASI cases stratified by gender, calendar year, and age group were then calculated. Finally, temporal trends were assessed by using the chi-square test for trend. For the inferential component of our study, we conducted unconditional logistic regression with SAS version 8.0 (SAS Institute, Cary, NC, USA). This test was initially performed by including in the model variables with a univariate likelihood ratio p value < 0.20. Among these factors, those with the highest multivariate Wald chi-square p value were then individually dropped, until the lowest Akaike Information Criterion value was attained. The McGill University Faculty of Medicine Institutional Review Board approved the study.

s with a univariate likelihood ratio p value < 0.20. Among these factors, those with the highest multivariate Wald chi-square p value were then individually dropped, until the lowest Akaike Information Criterion value was attained. The McGill University Faculty of Medicine Institutional Review Board approved the study. Results From 1995 through 2001, a total of 306 cases of IGASI were reported on the island of Montreal. The incidence of IGASI rose from 1.05 per 100,000 (19 cases) in 1995 to 1.71 (31 cases) in 1996 and 3.32 (60 cases) in 1997. After 1997, the incidence appeared to stabilize: 2.77 (50 cases) in 1998, 2.50 (45 cases) in 1999, 3.21 (58 cases) in 2000, and 2.37 (43 cases) in 2001. The average annual incidence of IGASI was 2.4 per 100,000. Most IGASI cases occurred in persons >40 years of age (172 [56%] of 306 cases) (Figure 1). The median age of patients was 46 years (range 1.5 months to 92 years). Figure 1 Annual incidence and death rate of invasive group A streptococcal infections, by age, in Montreal, Canada, 1995–2001.

Results From 1995 through 2001, a total of 306 cases of IGASI were reported on the island of Montreal. The incidence of IGASI rose from 1.05 per 100,000 (19 cases) in 1995 to 1.71 (31 cases) in 1996 and 3.32 (60 cases) in 1997. After 1997, the incidence appeared to stabilize: 2.77 (50 cases) in 1998, 2.50 (45 cases) in 1999, 3.21 (58 cases) in 2000, and 2.37 (43 cases) in 2001. The average annual incidence of IGASI was 2.4 per 100,000. Most IGASI cases occurred in persons >40 years of age (172 [56%] of 306 cases) (Figure 1). The median age of patients was 46 years (range 1.5 months to 92 years). Figure 1 Annual incidence and death rate of invasive group A streptococcal infections, by age, in Montreal, Canada, 1995–2001. Of the 306 reported IGASI cases, 112 (37%) were soft-tissue infections, 84 (28%) bacteremia, 32 (10%) pneumonia, and 29 (10%) STSS. Among patients with soft-tissue infections, 6 (5%) of 112 cases had myositis, 31 (28%) had cellulitis, and 76 (68%) had necrotizing fasciitis; 1 patient had both cellulitis and necrotizing fasciitis. We did not identify any significant trend over time with regard to the proportion of different clinical manifestations. As for specific clinical manifestations of IGASI, we estimated that bacteremia occurred, on average, in 0.66 per 100,000 persons each year, STSS in 0.23 per 100,000, soft-tissue infections in 0.89 per 100,000, and pneumonia in 0.25 per 100,000.

er time with regard to the proportion of different clinical manifestations. As for specific clinical manifestations of IGASI, we estimated that bacteremia occurred, on average, in 0.66 per 100,000 persons each year, STSS in 0.23 per 100,000, soft-tissue infections in 0.89 per 100,000, and pneumonia in 0.25 per 100,000. The predominant M serotypes included M1 (22%), M3 (12%), M28 (9%), M12 (8%), M4 (6%), and M6 (4%). Remaining serotypes accounted for <3% of isolates. Twenty percent of samples were nontypeable. Pneumonia The incidence of pneumonia appeared to significantly increase over time (χ2 = 5.65, p = 0.018), with an annual incidence of 0.06 per 100,000 in 1995 and 1996, 0.28 in 1997 and 1998, 0.39 in 1999, 0.50 in 2000, and 0.22 in 2001. This finding was confirmed by the odds of having pneumonia significantly increasing with each successive calendar year (adjusted odds ratio [aOR] = 1.21, 95% confidence interval [CI] 1.0–1.5). The proportion of women and girls with pneumonia (Figure 2) also significantly increased (χ2 = 5.03, p = 0.025), with women more likely to have pneumonia as compared to men (aOR 2.20, 95% CI 1.0–4.9). Gender was not associated with year in which the case occurred. Figure 2 Pneumonia as a proportion of invasive group A streptococcal infections (IGASI) by gender, Montreal, Canada, 1995–2001.

Pneumonia The incidence of pneumonia appeared to significantly increase over time (χ2 = 5.65, p = 0.018), with an annual incidence of 0.06 per 100,000 in 1995 and 1996, 0.28 in 1997 and 1998, 0.39 in 1999, 0.50 in 2000, and 0.22 in 2001. This finding was confirmed by the odds of having pneumonia significantly increasing with each successive calendar year (adjusted odds ratio [aOR] = 1.21, 95% confidence interval [CI] 1.0–1.5). The proportion of women and girls with pneumonia (Figure 2) also significantly increased (χ2 = 5.03, p = 0.025), with women more likely to have pneumonia as compared to men (aOR 2.20, 95% CI 1.0–4.9). Gender was not associated with year in which the case occurred. Figure 2 Pneumonia as a proportion of invasive group A streptococcal infections (IGASI) by gender, Montreal, Canada, 1995–2001. STSS We did not detect a significant secular trend in the occurrence of STSS (χ2 = 0.54, p = 0.46). Persons who abused alcohol (aOR 7.66, 95% CI 1.9–30.3]), were infected with serotype M9 (aOR 39.98, 95% CI 1.9–836]), or who had fasciitis (aOR 10.21, 95% CI 4.1–25.7]) were at a significantly greater risk of having STSS.

STSS We did not detect a significant secular trend in the occurrence of STSS (χ2 = 0.54, p = 0.46). Persons who abused alcohol (aOR 7.66, 95% CI 1.9–30.3]), were infected with serotype M9 (aOR 39.98, 95% CI 1.9–836]), or who had fasciitis (aOR 10.21, 95% CI 4.1–25.7]) were at a significantly greater risk of having STSS. Soft-tissue Infections No significant secular trend was apparent in the incidence of soft-tissue infections (χ2 = 0.48, p = 0.49). However, the odds of developing this manifestation as opposed to another significantly decreased with each successive calendar year (aOR 0.86, 95% CI 0.7–1.0). Drug use was weakly associated with soft-tissue infections (unadjusted OR 1.86, 95% CI 0.8–4.4). Given that trauma was a significant univariate risk marker for soft-tissue infections (OR 2.78, 95% CI 1.6–4.8), this association might have been attributable to injection drug use resulting in a trauma or wound. However, in our study, no correlation was seen between drug use and trauma (r = 0.002, p = 0.97). Furthermore, the association between drug use and soft-tissue infections became significant after adjusting for trauma or wound (OR 2.83, 95% CI 1.0–8.0). Varicella and serotypes M6, M12, or M22 were significant predictors for developing soft-tissue infections with aORs of 5.69 (95% CI 1.4–23.1), 4.3 (95% CI 1.1–16.7), 9.1 (95% CI 1.3–64.5), and 27.9 (95% CI 2.7–289), respectively. None of these factors were correlated with calendar year.

Soft-tissue Infections No significant secular trend was apparent in the incidence of soft-tissue infections (χ2 = 0.48, p = 0.49). However, the odds of developing this manifestation as opposed to another significantly decreased with each successive calendar year (aOR 0.86, 95% CI 0.7–1.0). Drug use was weakly associated with soft-tissue infections (unadjusted OR 1.86, 95% CI 0.8–4.4). Given that trauma was a significant univariate risk marker for soft-tissue infections (OR 2.78, 95% CI 1.6–4.8), this association might have been attributable to injection drug use resulting in a trauma or wound. However, in our study, no correlation was seen between drug use and trauma (r = 0.002, p = 0.97). Furthermore, the association between drug use and soft-tissue infections became significant after adjusting for trauma or wound (OR 2.83, 95% CI 1.0–8.0). Varicella and serotypes M6, M12, or M22 were significant predictors for developing soft-tissue infections with aORs of 5.69 (95% CI 1.4–23.1), 4.3 (95% CI 1.1–16.7), 9.1 (95% CI 1.3–64.5), and 27.9 (95% CI 2.7–289), respectively. None of these factors were correlated with calendar year. Bacteremia The incidence of bacteremia did not appear to change over time (χ2 = 0.56, p = 0.45). Only protective factors against bacteremia were identified: attending a school (aOR 0.15, 95% CI 0.0–0.7) and trauma or wound (aOR 0.4, 95% CI 0.2–0.9).

Varicella and serotypes M6, M12, or M22 were significant predictors for developing soft-tissue infections with aORs of 5.69 (95% CI 1.4–23.1), 4.3 (95% CI 1.1–16.7), 9.1 (95% CI 1.3–64.5), and 27.9 (95% CI 2.7–289), respectively. None of these factors were correlated with calendar year. Bacteremia The incidence of bacteremia did not appear to change over time (χ2 = 0.56, p = 0.45). Only protective factors against bacteremia were identified: attending a school (aOR 0.15, 95% CI 0.0–0.7) and trauma or wound (aOR 0.4, 95% CI 0.2–0.9). Death Due to IGASI The death ratio from IGASI was 15% (42 deaths among 306 cases). The highest proportion of known deaths was among patients with pneumonia (38%, 12 deaths among 32 pneumonia cases), followed by STSS (35%, 10 among 29), bacteremia (17%, 14 among 84), and soft-tissue infections (10%, 11 among 112). Within soft-tissue infections, necrotizing fasciitis had the highest risk for death among all age groups (16%, 5 deaths among 31 cases) followed by cellulitis and erysipelas (8%, 6 among 76). For myositis, among the 6 cases identified during a 7-year period, no deaths were recorded. No secular trends for death ratios were seen for any of the clinical manifestations of IGASI. Among those who died of IGASI, the most common serotypes were M1 (34%) and T1 (30%); however, neither was significantly associated with death (unadjusted OR for M1: 1.86, 95% CI 0.9–4.0; T1: 1.91, 95% CI 0.9–4.1). Predictors for death, after adjustment, are presented in the Table.

the clinical manifestations of IGASI. Among those who died of IGASI, the most common serotypes were M1 (34%) and T1 (30%); however, neither was significantly associated with death (unadjusted OR for M1: 1.86, 95% CI 0.9–4.0; T1: 1.91, 95% CI 0.9–4.1). Predictors for death, after adjustment, are presented in the Table. Table Adjusted odds ratio (OR) for factors associated with death attributable to invasive group A streptococcal infections, Montreal, Canada, 1995–2002 Variable OR (95% CI)* Age (y) 1.04 (1.0–1.1)† Underlying medical conditions No cancer Referent Cancer 4.14 (1.6–10.5) Primary site of infection Not cellulitis Referent Cellulitis 0.38 (0.1–1.0) Not pneumonia Referent Pneumonia 3.62 (1.4–9.0) Living environment Not working or living in hospital Referent Working or living in hospital 3.71 (1.0–13.6) M serotypes Not M2 Referent M2 10.69 (0.5–220) *CI, confidence interval. †Increase in risk per increase in year of age. Discussion When the results of our study are examined, several methodologic considerations must be taken into account. First, given that the administration of questionnaires for this study was not standardized, nondifferential misclassification could explain why certain factors in this study were not identified as potential markers for clinical manifestation outcomes. An additional limitation of our study was the low statistical power. For some measures of association, the probability of detecting a true association was estimated to be as low as 3%. As a result, while this study can identify potential predictors, it cannot exclude them.

Discussion When the results of our study are examined, several methodologic considerations must be taken into account. First, given that the administration of questionnaires for this study was not standardized, nondifferential misclassification could explain why certain factors in this study were not identified as potential markers for clinical manifestation outcomes. An additional limitation of our study was the low statistical power. For some measures of association, the probability of detecting a true association was estimated to be as low as 3%. As a result, while this study can identify potential predictors, it cannot exclude them. Additionally, given that this study was to a certain extent hypothesis-generating, some of the predictors found in this study (particularly those with weak associations) may have occurred by chance. Considering that an α level of 0.05 was used when testing ≈200 associations, at least 10 significant factors would be expected to be identified by chance. In our study, we identified 25 factors to be significantly associated with specific IGASI manifestations.

those with weak associations) may have occurred by chance. Considering that an α level of 0.05 was used when testing ≈200 associations, at least 10 significant factors would be expected to be identified by chance. In our study, we identified 25 factors to be significantly associated with specific IGASI manifestations. IGASI and STSS may be increasing in both incidence and severity (4). In particular, increasing trends in the IGASI incidence in the United States have been recorded in several hospital-based studies (29). Furthermore, past European studies noted a general increase in the incidence, although little evidence shows a trend occurring in the United States (14,30–32). While we documented a trend in the annual incidence of IGASI in Montreal during the first 3 years of our study, the incidence stabilized from 1997 onwards, which suggests that an initial rise in incidence might be attributable to underreporting immediately after IGASI became a notifiable disease. During the 7 years of our study, mortality did not appear to significantly change. Additionally, we could not identify any significant trends in the incidence and mortality of STSS. We did, however, ascertain that pneumonia attributable to IGASI significantly increased during 6 of the 7 years of our study. This finding was particularly evident among women. Our findings agree with those of a study in Ontario, which identified an increasing trend for pneumonia attributable to GAS from 1992 to 1999 (33).

id, however, ascertain that pneumonia attributable to IGASI significantly increased during 6 of the 7 years of our study. This finding was particularly evident among women. Our findings agree with those of a study in Ontario, which identified an increasing trend for pneumonia attributable to GAS from 1992 to 1999 (33). To the best of our knowledge, no research has been published on transmission rates for the different clinical manifestations of IGASI. However, the primary mode of person-to-person transmission of S. pyogenes is through respiratory droplets (5,6). Additionally, S. pyogenes that causes severe disease in one patient is more likely to cause severe disease in subsequent patients (6). Considering these previous study findings, one could hypothesize that secondary contacts of patients with respiratory manifestations might be more likely to acquire an infection leading to severe disease, compared to contacts of patients with other IGASI manifestations.

cause severe disease in subsequent patients (6). Considering these previous study findings, one could hypothesize that secondary contacts of patients with respiratory manifestations might be more likely to acquire an infection leading to severe disease, compared to contacts of patients with other IGASI manifestations. Even though IGASI is a reportable disease, our results for pneumonia may be an underestimate of the true values. Given that <1% of community-acquired pneumonia is attributed to S. pyogenes (34), pneumonia caused by this bacterium may have been ascribed to other causes and hence not reported. Our findings are further complicated by difficulties in defining pneumonia (33). No standard clinical definition distinguishes IGAS pneumonia from respiratory distress caused by STSS. Although both clinical manifestations might differ with regard to pathophysiology, given that prophylaxis is required for secondary contacts of either manifestation in Quebec, difficulties in distinguishing between these manifestations will probably not affect the public health implications of our findings.

ed by STSS. Although both clinical manifestations might differ with regard to pathophysiology, given that prophylaxis is required for secondary contacts of either manifestation in Quebec, difficulties in distinguishing between these manifestations will probably not affect the public health implications of our findings. With regard to the generalizability of our results, when comparing our findings with previously published studies, we did not detect any geographic differences in the incidence of IGASI (17). Our data showed that the yearly incidence of IGASI in Montreal (1.0–3.3 per 100,000) was similar to the incidence of IGASI in British Columbia (20), Ontario (16), Israel (35), Sweden (19,22), and the United States (14,18). Furthermore, death rates from IGASI in Montreal were comparable to death rates calculated for British Colombia (20) and Sweden (19,22). Only Arizona appeared to have a higher death rate due to IGASI, at 20% (3). This difference might be attributable to the elevated prevalence of diabetes (a risk factor for IGASI) in the Arizona community studied (3).

ASI in Montreal were comparable to death rates calculated for British Colombia (20) and Sweden (19,22). Only Arizona appeared to have a higher death rate due to IGASI, at 20% (3). This difference might be attributable to the elevated prevalence of diabetes (a risk factor for IGASI) in the Arizona community studied (3). Along with these descriptive findings, we identified several factors associated with clinical manifestations of IGAS and associated death. Having varicella before IGASI increased the risk of developing a soft-tissue infection 6 times and the risk of dying 5 times. Although we could not identify any literature linking varicella infection with soft-tissue infections, given that soft-tissue infections are the predominant clinical manifestation of IGASI, our findings support previous research that suggests that varicella might be an important risk factor for developing IGASI (16,36). Soft-tissue infections were almost twice as likely to develop in persons using drugs. This association could be attributable to injection drug use; however, it remains even after controlling for trauma. One explanation for this unexpected finding could be nondifferential misclassification. A subanalysis of drug use showed that 23% of patients indicated a trauma or wound. However, we were unable to determine the reliability of reporting. While a patient could have affirmatively answered to drug use, a wound inflicted by intravenous drug use may not have been considered sufficiently severe to indicate a trauma or wound.

of drug use showed that 23% of patients indicated a trauma or wound. However, we were unable to determine the reliability of reporting. While a patient could have affirmatively answered to drug use, a wound inflicted by intravenous drug use may not have been considered sufficiently severe to indicate a trauma or wound. Our descriptive analysis appears to support previous research findings that those <1 year of age and those >60 years of age have the highest incidence of IGASI (3,21–23). However, previous studies also suggest that children might have a lower incidence of STSS and be at a decreased risk of dying of IGASI (21). This research includes a study that identified a nonsignificant 5-fold rise in risk for death per year increase in age. In contrast, our study showed a 2%–4% increase. Our study finding that M1 and M3 accounted for >30% of all isolates tested for M surface proteins was consistent with previous studies that reported these 2 M serotypes as the most common for IGASI (3,15,16,23,24). This finding is also consistent with the choice of serotypes to include in streptococcal vaccines being evaluated at the moment. The hexavalent vaccine (37) is composed of serotypes 1, 3, 5, 6, 19, and 24; these types represent 38% of isolates in our study. The types in the 26-valent vaccine (38) represent 70% of our isolates. This number includes M1 with 22%; M3 with 12%; M28 with 9%; M12 with 8%; M6 with 4%; M22 and M11 each with 3%; M89 with 2%; and M75, M2, M77, M43, M5, M76, and M33 each with 1%.

, 5, 6, 19, and 24; these types represent 38% of isolates in our study. The types in the 26-valent vaccine (38) represent 70% of our isolates. This number includes M1 with 22%; M3 with 12%; M28 with 9%; M12 with 8%; M6 with 4%; M22 and M11 each with 3%; M89 with 2%; and M75, M2, M77, M43, M5, M76, and M33 each with 1%. Furthermore, our study confirmed univariate model findings from a study by O’Brien et al. that found M3 to be significantly associated with STSS (14). However, this association did not remain in our multivariate analysis. Our findings would thus appear to concur with those of a another (case-control) study that found while M1 and M3 may be significant risk factors for IGASI, once a person is infected, environmental and host factors might have a role in determining the type of invasive disease manifestations (8,9). This finding could explain why IGASI may develop in patients infected with the same strain of GAS but have different clinical manifestations of the disease (e.g., STSS versus pneumonia) (9). Future epidemiologic studies of risk factors for clinical manifestations of IGASI might be designed to look at risk factors separately in patients identified with M1 and M3 serotypes. By doing so, nondifferential misclassification might be minimized and risk factors with weaker associations might be more easily identified.

re epidemiologic studies of risk factors for clinical manifestations of IGASI might be designed to look at risk factors separately in patients identified with M1 and M3 serotypes. By doing so, nondifferential misclassification might be minimized and risk factors with weaker associations might be more easily identified. Suggested citation for this article: Hollm-Delgado M-G, Allard R, Pilon PA. Group A streptococcal infections, manifestations and their predictors, Montreal, 1995–2002. Emerg Infect Dis [serial on the Internet]. 2005 Jan [date cited]. http://dx.doi.org/10.3201/eid1101.030651 Acknowledgments We thank Lucie Bédard for providing permission to use the database to conduct this study and for her helpful comments on an initial draft of the protocol, Louise Marcotte for preparing the dataset, Gilles Paradis for his helpful comments on an initial draft of this paper, Paul Rivest for being available for questions regarding the surveillance procedure, and Agnès Beaume for her administrative assistance. Ms. Hollm-Delgado is a doctoral student in public health at the Université de Montréal. This paper originates from research she completed at the Direction de Santé Publique (Montreal, Canada) while earning her Masters degree in epidemiology from McGill University. Her current research interests include respiratory infections and antimicrobial resistance.

Worldwide, childhood deaths have decreased, largely attributable to fewer deaths from pneumonia, measles, and diarrhea (1); some of these reductions have been achieved through vaccination against common bacterial pathogens such as Streptococcus pneumoniae and Haemophilus influenzae type b (2). However, progress in reducing deaths among children has been slower in sub-Saharan Africa, where approximately half of all such deaths occur, a third during the first month of life (1). To achieve further disease reductions, it is essential to address other, potentially preventable, causes of invasive bacterial disease, such as group A Streptococcus (GAS). It is estimated that >660,000 cases of invasive GAS infection occur each year; >95% cases occur in resource-poor regions, and >160,000 patients die (3). Despite these estimates, data on invasive GAS infections in resource-poor settings are limited.

ses of invasive bacterial disease, such as group A Streptococcus (GAS). It is estimated that >660,000 cases of invasive GAS infection occur each year; >95% cases occur in resource-poor regions, and >160,000 patients die (3). Despite these estimates, data on invasive GAS infections in resource-poor settings are limited. The Young Infant Study of invasive bacterial disease conducted in the late 1990s in The Gambia, Ethiopia, Papua New Guinea, and the Philippines reported GAS in 29 (17%) of 167 bacterial isolates from blood cultures and in 3 (7.5%) of 40 cerebrospinal fluid (CSF) cultures (4). Although this finding meant that GAS was the third most commonly isolated bacterium after S. pneumoniae and Staphylococcus aureus, research into associated invasive GAS infections has been limited. To our knowledge, in sub-Saharan Africa, only 1 estimate of invasive GAS incidence has been published: 29 cases/100,000 person-years (definite cases of bacteremia only) among children <5 years of age in Kenya and 96 cases/100,000 person-years among children <1 year of age (5). These incidences are higher than those reported from other resource-poor settings. Data from Fiji, in the Pacific, report an incidence of 26 cases/100,000 person-years among children <5 years and 45 cases/100,000 person-years among children <1 year of age (6). In New Caledonia, the incidence for children <5 years of age was 7 cases/100,000 person-years (7).

reported from other resource-poor settings. Data from Fiji, in the Pacific, report an incidence of 26 cases/100,000 person-years among children <5 years and 45 cases/100,000 person-years among children <1 year of age (6). In New Caledonia, the incidence for children <5 years of age was 7 cases/100,000 person-years (7). Vaccines for GAS are being developed; the most advanced is a 30-valent serotype-specific vaccine. Data about the emm types causing invasive GAS disease in sub-Saharan Africa are critical for assessing potential vaccine serotype coverage. Through comprehensive prospective clinical and microbiological surveillance (1998–2011), we determined incidence, clinical characteristics, and outcomes among children with invasive GAS infections in a hospital in rural Kenya. We used whole-genome sequencing to determine emm types and phylogenetic variations of invasive GAS isolates.

ensive prospective clinical and microbiological surveillance (1998–2011), we determined incidence, clinical characteristics, and outcomes among children with invasive GAS infections in a hospital in rural Kenya. We used whole-genome sequencing to determine emm types and phylogenetic variations of invasive GAS isolates. Materials and Methods Study Design and Participants Since 1998, the Kenya Medical Research Institute/Wellcome Trust Research Programme has undertaken prospective systematic clinical surveillance, including standardized clinical documentation and systematic microbiological investigation, for invasive bacterial disease among all children admitted for medical care to Kilifi County Hospital (in Kilifi, a rural area of coastal Kenya), as described elsewhere (5,8). Our observational study identified cases of invasive GAS disease during this surveillance of all children admitted to Kilifi County Hospital from August 1, 1998, through December 31, 2011. The study size was determined by admissions during the study period. The study was approved by the National Ethics Committee, Nairobi, Kenya (ERC 2144), and the Oxford Tropical Research Ethics Committee (OXTREC 151–12). The denominator population was determined by using the Kilifi Health and Demographic Surveillance System, which covers 891 km2 surrounding the hospital and in 2011 included ≈260,000 residents (9); household enumerations are performed quarterly. We calculated the population age structure at the midpoint of the study (mid-2004) and the total number of live births.

using the Kilifi Health and Demographic Surveillance System, which covers 891 km2 surrounding the hospital and in 2011 included ≈260,000 residents (9); household enumerations are performed quarterly. We calculated the population age structure at the midpoint of the study (mid-2004) and the total number of live births. Clinical Surveillance and Case Definitions At the time of patient admission to the hospital, a standardized set of clinical symptoms and signs were recorded and prospectively entered into a database. At the time of patient discharge, outcome was recorded. Anthropometry for the presence of kwashiorkor (edematous malnutrition) was systematically undertaken at admission and used to define severe acute malnutrition (10). For all nonelective admissions, samples were collected for complete blood count, malaria slide, and blood culture. If clinically indicated, culture was performed for CSF, urine, and pus swab samples. Inpatient treatment was provided according to World Health Organization guidelines (10).

severe acute malnutrition (10). For all nonelective admissions, samples were collected for complete blood count, malaria slide, and blood culture. If clinically indicated, culture was performed for CSF, urine, and pus swab samples. Inpatient treatment was provided according to World Health Organization guidelines (10). Starting in January 2007, in line with national guidelines, HIV testing by rapid test was offered for all children admitted. For children who had invasive GAS disease before 2007 and were not tested during our previous study of bacteremia (5), a trained counselor visited households and offered voluntary counseling and testing (9). For children who had died or were untraceable, a stored blood sample was tested by PCR for HIV. Sickle cell disease testing by electrophoresis was undertaken as clinically indicated; for children admitted with bacteremia during 1998–2008, PCR was used to retrospectively test for sickle cell disease, as described previously (11). For this analysis, data were extracted from clinical and laboratory databases. All paper clinical records were reviewed for signs and symptoms relevant to invasive GAS disease, including the presence of pharyngitis, burns, scabies, and a vesicular rash suggestive of varicella or herpes zoster infection.

Starting in January 2007, in line with national guidelines, HIV testing by rapid test was offered for all children admitted. For children who had invasive GAS disease before 2007 and were not tested during our previous study of bacteremia (5), a trained counselor visited households and offered voluntary counseling and testing (9). For children who had died or were untraceable, a stored blood sample was tested by PCR for HIV. Sickle cell disease testing by electrophoresis was undertaken as clinically indicated; for children admitted with bacteremia during 1998–2008, PCR was used to retrospectively test for sickle cell disease, as described previously (11). For this analysis, data were extracted from clinical and laboratory databases. All paper clinical records were reviewed for signs and symptoms relevant to invasive GAS disease, including the presence of pharyngitis, burns, scabies, and a vesicular rash suggestive of varicella or herpes zoster infection. Cases of invasive GAS were defined as definite if GAS was isolated from a normally sterile site (blood, CSF, or other sterile fluid/tissue) or if necrotizing fasciitis with evidence of GAS infection was present (e.g., typical gram-positive cocci found after Gram staining or serologic testing results positive for streptococci). Cases of invasive GAS were defined as probable if any of the following were found: classic necrotizing fasciitis without microbiological confirmation; cellulitis in a patient who was moderately or severely unwell (i.e., unwell and history of parenteral receipt of antimicrobial drugs, admission to hospital, or both); microbiological confirmation (i.e., growth of GAS on culture of swab sample or serologic test results positive for streptococci); or other clinically relevant infection in a patient who is moderately or severely unwell (i.e., unwell and history of parenteral receipt of antimicrobial drugs, admission to hospital, or both), in conjunction with positive GAS culture from deep wound swab sample or biopsy sample from surgical infection site (6).

occi); or other clinically relevant infection in a patient who is moderately or severely unwell (i.e., unwell and history of parenteral receipt of antimicrobial drugs, admission to hospital, or both), in conjunction with positive GAS culture from deep wound swab sample or biopsy sample from surgical infection site (6). Clinical syndromes of invasive GAS disease vary. These syndromes were categorized as meningitis, severe pneumonia, skin or soft tissue infection, joint and bone infection, necrotizing fasciitis, urinary tract infection, acute glomerulonephritis, abdominal disease, endocarditis, bacteremia with no focus, and streptococcal toxic shock syndrome (Technical Appendix Table 1). Microbiological and Molecular Methods Blood cultures were undertaken by using the BACTEC Peds Plus system (Becton Dickinson, Franklin Lakes, NJ, USA) according to the manufacturers’ instructions. Positive broth cultures and CSF, urine, and surface swab samples were subcultured on 5% horse blood agar and chocolate agar. GAS isolates were identified by β-hemolysis, followed by Gram staining and catalase testing, and then grouped by latex bead agglutination. Penicillin susceptibility was tested by disk diffusion (http://www.bsac.org.uk/). Laboratory procedures were subject to internal quality control and external quality control by the UK National External Quality Assessment Service.

lowed by Gram staining and catalase testing, and then grouped by latex bead agglutination. Penicillin susceptibility was tested by disk diffusion (http://www.bsac.org.uk/). Laboratory procedures were subject to internal quality control and external quality control by the UK National External Quality Assessment Service. GAS isolates were subcultured on 5% horse blood agar from archived bacterial isolates (stored at −80°C) and transported to the Wellcome Trust Sanger Institute, Cambridge, UK. DNA was extracted by a QIAxtractor (QIAGEN, Valencia, CA, USA), and DNA quality and quantity were documented by using NanoDrop (Thermo Scientific, Waltham, MA, USA) and Qubit (Life Technologies, Carlsbad, CA, USA) techniques. Whole-genome sequences were determined from Illumina 96-plex libraries by using the HiSeq2000 sequencing platform (Illumina, San Diego, CA, USA) to generate tagged 75-bp paired-end reads. To obtain the overall population structure of the sequenced genomes, we mapped individual Illumina read pairs to the MGAS5005 (emm1) reference genome (12) by using SMALT version 0.5.8 (http://www.sanger.ac.uk/resources/software/smalt/). The average coverage of the resulting whole-genome alignment was 190×. The minimum base-call quality for identifying a single nucleotide polymorphism (SNP) was set at 50, and the minimum mapping quality for SNP calling was set at 30. SNPs called in known MGAS5005 prophage regions and repeat regions were excluded from analyses. The final genome alignment was 1,629,062 bp and comprised 125,233 SNPs. To examine the genomic relationships between the sequenced genomes, we generated a maximum-likelihood tree from the SNP alignment by using FastTree (13). Draft genome assemblies were compiled by using an iterative sequence assembly process as defined previously (14). An initial quality control screen of the short-read sequences to identify mixed isolates and low-quality sequences was determined by examining genome assembly length and SNP heterogeneity. A total of 43 (11.6%) sequences had an assembly length of >2 mega basepairs and were excluded from phylogenetic analyses because of possible contamination. The emm type and multilocus sequence type (MLST) were obtained from in-house BLAST analysis of draft genome assemblies and compared with those in centralized databases (http://www.cdc.gov/streplab/m-proteingene-typing.html, http://pubmlst.org/spyogenes/). New emm and MLST alleles were assigned by database curators.

on. The emm type and multilocus sequence type (MLST) were obtained from in-house BLAST analysis of draft genome assemblies and compared with those in centralized databases (http://www.cdc.gov/streplab/m-proteingene-typing.html, http://pubmlst.org/spyogenes/). New emm and MLST alleles were assigned by database curators. Allocation of emm cluster was derived as previously described (7). Heterogeneity observed within the typing schemes was investigated by using maximum-likelihood associations in whole-genome sequence data. Epidemiologic Analysis Epidemiologic analyses were undertaken by using STATA version 13 statistical software (StataCorp LP, College Station, TX, USA). Clinical characteristics of children with invasive GAS disease were tabulated, and the frequency of clinical syndromes of invasive GAS disease and associated case-fatality risks (by age group) were calculated. Incidence rates were calculated by using the invasive GAS cases resident within the Kilifi Health and Demographic Surveillance System, the age structure of the population at the study mid-point (2004), and the total number of live births. Trends in admissions were examined by using rolling averages, and a comparison between seasons (wet and dry) was made by using the Poisson distribution.

t within the Kilifi Health and Demographic Surveillance System, the age structure of the population at the study mid-point (2004), and the total number of live births. Trends in admissions were examined by using rolling averages, and a comparison between seasons (wet and dry) was made by using the Poisson distribution. Results During the study, 64,761 children were admitted to the hospital with acute illness. From 370 children with invasive GAS infection, 391 GAS isolates were identified. Of these 391 isolates, 154 (39.4%) were from blood, 9 (2.3%) from CSF, 214 (54.7%) from a swab sample (wound, skin breach, or pus), 8 (2.0%) from joint aspirates, and 6 (1.5%) from urine. From 20 children, >1 GAS isolate was identified: 7 children had invasive GAS isolated from both blood and CSF; 2 children had repeat positive blood cultures; 2 children had invasive GAS isolated from blood and a swab sample; 1 child had invasive GAS isolated from CSF and a swab sample; 7 children had invasive GAS isolated from 2 swab samples; and 1 child had invasive GAS isolated from 3 swab samples. No isolates were resistant to penicillin.

repeat positive blood cultures; 2 children had invasive GAS isolated from blood and a swab sample; 1 child had invasive GAS isolated from CSF and a swab sample; 7 children had invasive GAS isolated from 2 swab samples; and 1 child had invasive GAS isolated from 3 swab samples. No isolates were resistant to penicillin. Characteristics of Children and Risk Factors for Definite Invasive GAS Disease Full clinical information was available for 369 of the 370 children: 152 children had definite and 217 had probable invasive GAS disease as defined. A total of 94 (25.5%) cases of invasive GAS were in neonates (Table 1). Among the 152 children with definite invasive GAS disease, 5 (3.3%) had burns, 4 (2.6%) had concurrent scabies, 1 (0.7%) had a vesicular rash (consistent with herpes zoster or varicella), and 2 (1.3%) had a history of trauma. Among the 217 with probable invasive GAS disease, 26 (12.0%) had burns, 3 (1.4%) had scabies, 1 (0.5%) had a vesicular rash, and 4 (1.8%) had a history of trauma (risk factors were not mutually exclusive). No reports of pharyngitis were documented for patients who had definite or probable invasive GAS disease. Among the 152 children with definite invasive GAS disease, prevalence of common risk factors for invasive bacterial disease was high: 81 (53.3%) had any risk factor; 30 (19.7%) had severe acute malnutrition, including 9 (5.9%) with kwashiorkor; 28 (18.4%) had malaria (slide positive for Plasmodium falciparum), and 24 (15.8%) had HIV infection.

th definite invasive GAS disease, prevalence of common risk factors for invasive bacterial disease was high: 81 (53.3%) had any risk factor; 30 (19.7%) had severe acute malnutrition, including 9 (5.9%) with kwashiorkor; 28 (18.4%) had malaria (slide positive for Plasmodium falciparum), and 24 (15.8%) had HIV infection. Table 1 Characteristics of children with GAS disease admitted to Kilifi County Hospital, Kenya, 1998–2011* Characteristic All GAS disease, n = 369, no. (%) Definite invasive GAS disease, n = 152, no. (%) Probable invasive GAS disease, n = 217, no. (%) Age 0–6 d 33 (8.9) 13 (8.6) 20 (9.2) 7–28 d 61 (16.5) 38 (25.0) 23 (10.6) 29–59 d 17 (4.6) 12 (7.9) 5 (2.3) 60 d–1 y 63 (17.1) 40 (26.3) 23 (10.6) >1 and <5 y 125 (33.9) 41 (27.0) 84 (38.7) 5–12 y 70 (19.0) 8 (5.3) 62 (28.6) Sex M 219 (59.3) 84 (55.3) 135 (62.2) F 150 (40.7) 68 (44.7) 82 (37.8) Severe acute malnutrition No 294 (79.7) 106 (69.7) 188 (86.6) Yes (wasting) 47 (12.7) 30 (19.7) 17 (7.8) Yes (kwashiorkor) 11 (3.0) 9 (5.9) 2 (0.9) Not known 17 (4.6) 7 (4.6) 10 (4.6) Malaria (positive slide result) No 313 (84.8) 123 (80.9) 190 (87.6) Yes 56 (15.2) 29 (19.1) 27 (12.4) HIV infection No 209 (56.6) 116 (76.3) 93 (42.9) Yes 28 (7.6) 24 (15.8) 4 (1.8) Not known 132 (35.8) 12 (7.9) 120 (55.3) Sickle cell disease No 136 (36.9) 95 (62.5) 41 (18.9) Sickle cell trait 14 (3.8) 9 (5.9) 5 (2.3) Sickle cell disease 3 (0.8) 1 (0.7) 2 (0.9) Not known 216 (58.5) 47 (30.9) 169 (77.9) *Malaria incidence (slide-positive admissions data from Kilifi Health and Demographic Surveillance System) decreased from 28.5 to 3.45 cases per 1,000 person-years during 1999–2007. HIV prevalence was 4.9% (routine antenatal screening, 2004–2007) with no evidence of a temporal trend. Sickle cell disease prevalence among infants in the Kilifi Health and Demographic Surveillance System (2006–2009) was 15% for genotypes HbAS and 1% with HbSS (11). Severe acute malnutrition is referenced against World Health Organization population standards (Technical Appendix Table 1). GAS, group A Streptococcus.

nd. Sickle cell disease prevalence among infants in the Kilifi Health and Demographic Surveillance System (2006–2009) was 15% for genotypes HbAS and 1% with HbSS (11). Severe acute malnutrition is referenced against World Health Organization population standards (Technical Appendix Table 1). GAS, group A Streptococcus. Clinical Syndromes of GAS Disease and Case-Fatality Risk Among the 369 children with invasive GAS disease, the most frequent infection was skin or soft tissue infection, occurring in 258 (69.9%); followed by severe pneumonia in 86 (23.3%), of which 59 (69%) were complicated by sepsis; then bacteremia without focus in 53 (14.4%) (Table 2). Also among these 369 children, 17 (4.6%) had bone and joint infections, 11 (3.0%) had meningitis, 6 (1.6%) had a urinary tract infection, 2 (0.5%) had acute glomerulonephritis, 1 (0.3%) had endocarditis, 1 (0.3%) had nonspecific abdominal signs, and 1 (0.3%) had necrotizing fasciitis. A total of 19 (5.1%) cases met the criteria for streptococcal toxic shock syndrome (15). Of the 369 children, 45 (12.2%) died. The case-fatality risk was highest among those with severe pneumonia (20/86, 23.3%), followed by primary bacteremia (11/53, 20.8%) and meningitis (2/11, 18.2%). Pneumonia and primary bacteremia occurred most frequently among children <1 year of age.

occal toxic shock syndrome (15). Of the 369 children, 45 (12.2%) died. The case-fatality risk was highest among those with severe pneumonia (20/86, 23.3%), followed by primary bacteremia (11/53, 20.8%) and meningitis (2/11, 18.2%). Pneumonia and primary bacteremia occurred most frequently among children <1 year of age. Table 2 Common clinical syndromes of GAS disease among children admitted to Kilifi County Hospital, Kenya, 1998–2011 Clinical syndrome Age 0–6 d 7–28 d 29–59 d 60 d–1 y >1–<5 y 5–12 y Overall All cases No. (%) 33 (100) 61 (100) 17 (100) 63 (100) 125 (100) 70 (100) 369 (100) Deaths, CFR 10 (30.3) 23 (37.7) 1 (6.3) 7 (11.1) 14 (11.2) 1 (1.4) 45 (12.2) Skin and soft tissue infection No. (%) 22 (66.7) 33 (54.1) 5 (29.4) 37 (58.7) 99 (79.2) 62 (88.6) 258 (69.9) Deaths, CFR 6 (27.3) 4 (12.1) 0 1 (2.7) 6 (6.1) 1 (1.6) 17 (4.5) Severe pneumonia† No. (%) 7 (21.2) 17 (27.9) 8 (47.1) 28 (44.4) 21 (16.8) 3 (4.3) 86 (23.3) Deaths, CFR 2 (28.6) 5 (29.4) 0 5 (17.9) 8 (38.1) 0 20 (23.3) Primary bacteremia No. (%) 8 (24.2) 17 (27.9) 3 (17.6) 9 (14.3) 13 (10.4) 3 (4.3) 53 (14.4) Deaths, CFR 2 (25.0) 5 (29.4) 1 (33.3) 2 (22.2) 1 (7.7) 0 11 (20.8) *CFR, case-fatality risk; GAS, group A Streptococcus. †59 of the 86 severe pneumonia cases were complicated by sepsis.

5 (29.4) 0 5 (17.9) 8 (38.1) 0 20 (23.3) Primary bacteremia No. (%) 8 (24.2) 17 (27.9) 3 (17.6) 9 (14.3) 13 (10.4) 3 (4.3) 53 (14.4) Deaths, CFR 2 (25.0) 5 (29.4) 1 (33.3) 2 (22.2) 1 (7.7) 0 11 (20.8) *CFR, case-fatality risk; GAS, group A Streptococcus. †59 of the 86 severe pneumonia cases were complicated by sepsis. Incidence of Invasive GAS Disease The minimum incidence (cases/100,000 person-years) for definite and all (definite and probable) invasive GAS disease, respectively, among children <5 years of age was 17 (95% CI 14–21) and 35 (95% CI 30–40); among children <1 year of age, incidence was 59 (95% CI 45–74) and 101 (95% CI 83–121). Among neonates, incidence (cases/1,000 live births) for definite and all invasive GAS, respectively, was 0.3 (95% CI 0.2–0.4) and 0.6 (95% CI 0.4–0.7). The incidence of death was 0.1 (95% CI 0.1–0.2) deaths per 1,000 live births (Table 3). No trend was detected in the number of cases admitted over the study period (Technical Appendix Figure 1). Invasive GAS cases occurred less frequently during the dry months across all years (December–March, 26 cases/month) than during months of the short and long rains (April–October, 33 cases/month) (p = 0.029).

e 3). No trend was detected in the number of cases admitted over the study period (Technical Appendix Figure 1). Invasive GAS cases occurred less frequently during the dry months across all years (December–March, 26 cases/month) than during months of the short and long rains (April–October, 33 cases/month) (p = 0.029). Table 3 Estimated minimum incidence of definite and probable invasive GAS disease and deaths associated with invasive GAS disease in the catchment area of Kilifi County Hospital, Kenya, 1998–2011* Incidence† Age group Neonate, 0–27 d, n = 9,828‡ Infant, 28–59 d, n = 10,463‡ Infant, 2–11 mo, n = 92,070‡ Child 1–4 y, n = 453,857‡ Child 5–12 y, n = 730,512‡ Probable and definite invasive GAS disease incidence (95% CI) 631 (484–808) 105 (52–188) 43 (31–59) 19 (15–23) 6 (4–9) Definite invasive GAS disease incidence (95% CI) 326 (223–459) 86 (39–163) 27 (18–40) 7 (5–10) 1 (0–1) Death associated with all invasive GAS disease (95% CI) 163 (93–264) 10 (0–53) 5 (2–13) 2 (1–3) 0 (0–1) *GAS, group A Streptococcus. †Per 100,000 person-years. ‡Population denominator in person-years.

(15–23) 6 (4–9) Definite invasive GAS disease incidence (95% CI) 326 (223–459) 86 (39–163) 27 (18–40) 7 (5–10) 1 (0–1) Death associated with all invasive GAS disease (95% CI) 163 (93–264) 10 (0–53) 5 (2–13) 2 (1–3) 0 (0–1) *GAS, group A Streptococcus. †Per 100,000 person-years. ‡Population denominator in person-years. Molecular Epidemiology of GAS Of the 391 original GAS isolates, we retrieved 371 and generated high-quality genome sequences for 328 (Technical Appendix Table 2). From another 29 GAS isolates (combined total of 357) with lower quality genome sequences, we were able allocate an emm type. The remaining 14 samples were subsequently excluded from molecular analyses because they were not GAS or were mixed cultures, affecting accurate SNP calling (but not epidemiologic analyses because these isolates had been subcultured, stored, and then subcultured again, potentially introducing contamination). Through BLAST analysis of the 357 genome sequences against the emm typing database, we assigned 88 different emm types (97 including subtypes). Of the emm subtypes, 21 were new variants. No emm types represented >5% of the isolates studied, showing that no single emm type was predominant in the GAS population irrespective of clinical association (Figure 1, Technical Appendix Figure 2).

yping database, we assigned 88 different emm types (97 including subtypes). Of the emm subtypes, 21 were new variants. No emm types represented >5% of the isolates studied, showing that no single emm type was predominant in the GAS population irrespective of clinical association (Figure 1, Technical Appendix Figure 2). Figure 1 emm types of group A Streptococcus (GAS) isolates from children with GAS disease admitted to Kilifi County Hospital, Kenya, 1998–2011. emm types shown in green are included in the 30-valent vaccine; emm types in blue are not included in the 30-valent vaccine, but this vaccine may provide immunity to this emm type through cross-reactivity; emm types in red are not included in the 30-valent vaccine, and there is no evidence of cross-reactivity; emm types in yellow are not included in the 30-valent vaccine, and their cross-reactivity has not yet been tested.

30-valent vaccine, but this vaccine may provide immunity to this emm type through cross-reactivity; emm types in red are not included in the 30-valent vaccine, and there is no evidence of cross-reactivity; emm types in yellow are not included in the 30-valent vaccine, and their cross-reactivity has not yet been tested. Of the 357 GAS isolates, we assigned an emm cluster designation to 329 on the basis of the recently described emm cluster classification scheme (16). Of the 48 emm clusters described, 24 were represented within the Kilifi invasive GAS population of isolates (Technical Appendix Table 3). Of the 140 MLSTs identified, only 24 sequence types were represented within the MLST database (78/328 strains with high-quality whole-genome sequence data). We identified 89 new allelic variants among the 7 housekeeping genes and assigned 116 new MLSTs. Crude phylogenetic analyses of the Kilifi invasive GAS population as a whole revealed a star-like topology (Figure 2) indicative of diverse core genotypes. Collectively, these data illustrate substantial heterogeneity within invasive GAS genotypes in the Kilifi population.

housekeeping genes and assigned 116 new MLSTs. Crude phylogenetic analyses of the Kilifi invasive GAS population as a whole revealed a star-like topology (Figure 2) indicative of diverse core genotypes. Collectively, these data illustrate substantial heterogeneity within invasive GAS genotypes in the Kilifi population. Figure 2 Population structure of 328 Streptococcus pyogenes strains from children with group A Streptococcus (GAS) disease admitted to Kilifi County Hospital, Kenya, 1998–2011. Unrooted maximum-likelihood phylogeny based on the whole-genome associations of mapped S. pyogenes genomes to the MGAS5005 reference genome indicates extensive genomic diversity within the population. The rings surrounding the central phylogeny correspond to standard GAS molecular typing methods; colors indicate different STs. Inner ring, emm ST (16); middle ring, emm cluster (17); outer ring, multilocus sequence type (18). NT, nontypeable emm clusters; ST, sequence type. *Position of the MGAS5005 reference genome. Scale bar indicates genetic change of 0.01. In terms of vaccine coverage, 99 (28%) of 357 GAS isolates are included within the current 30-valent vaccine (19), and another 104 (29%) exhibit a degree of emm cross-reactivity in vitro (Figure 1) (20). Of the remainder, 27 (8%) were not included in the vaccine and are not cross-reactive, and 127 (36%) have not yet been investigated for cross-reactivity.

) of 357 GAS isolates are included within the current 30-valent vaccine (19), and another 104 (29%) exhibit a degree of emm cross-reactivity in vitro (Figure 1) (20). Of the remainder, 27 (8%) were not included in the vaccine and are not cross-reactive, and 127 (36%) have not yet been investigated for cross-reactivity. Discussion Incidence of invasive GAS disease in this rural sub-Saharan African setting was strikingly high, particularly among children in the first year of life among whom GAS was a major cause of sepsis and severe pneumonia. The minimum incidence of invasive GAS infection was highest among neonates (0.6 cases/1,000 live births; more than one third of all case-patients died). Minimum incidence in the first year of life overall was also high (101 cases/100,000 person-years), twice that for Fiji, the only other resource-poor setting from which an incidence estimate is available (6). The incidence estimates presented here are probably underestimates because inclusion in the study relied on hospital admission; hence, they are referred to as minimum incidence estimates. Residents living nearer to Kilifi County Hospital are more likely to access care than those living farther from it (21), and care-seeking behavior varies (22). The incidence of invasive GAS is probably accompanied by high prevalence of the spectrum of GAS infections, including acute poststreptococcal glomerulonephritis and acute rheumatic fever, which can lead to rheumatic heart disease (23); however, data for sub-Saharan Africa are limited (24,25).

king behavior varies (22). The incidence of invasive GAS is probably accompanied by high prevalence of the spectrum of GAS infections, including acute poststreptococcal glomerulonephritis and acute rheumatic fever, which can lead to rheumatic heart disease (23); however, data for sub-Saharan Africa are limited (24,25). In rural Kenya, unlike in other settings, pharyngitis, varicella, and scabies did not seem to be major drivers of invasive GAS disease (23,26), and impetigo was not differentiated from skin infections. These conditions are probably underascertained because they would not in themselves result in hospital admission, and unlike most of the clinical and microbiological data (systematically sought and collected), these diagnoses relied on observations being recorded. Also, despite the high frequency of skin and soft tissue infections, we detected only 1 case of necrotizing fasciitis, which may again be underascertainment from clinical information. Invasive GAS was, however, associated with concurrent conditions driving other bacterial diseases in sub-Saharan Africa: HIV, severe acute malnutrition, and malaria (5,27,28) but not sickle cell disease (as reported elsewhere) (11,29–31).

izing fasciitis, which may again be underascertainment from clinical information. Invasive GAS was, however, associated with concurrent conditions driving other bacterial diseases in sub-Saharan Africa: HIV, severe acute malnutrition, and malaria (5,27,28) but not sickle cell disease (as reported elsewhere) (11,29–31). In this study, the invasive GAS emm types and emm clusters were extremely heterogeneous and differed from those that cause disease in resource-rich settings. The presence of several S. dysgalactiae subsp. equisimilis–like emm types within a S. pyogenes genomic backbone supports previous observations of interspecies genetic transfer of emm alleles (32). The overall diversity of emm types we describe supports findings of increased heterogeneity in other resource-poor settings (33). One published study reports noninvasive GAS emm types from sub-Saharan Africa. In that study, school children in Ethiopia were investigated for GAS carriage; 43 different emm types were identified in 82 colonizing GAS isolates (34). Less than one third of emm types identified in our study were also identified in the Ethiopia study, suggesting that the pool of GAS emm types in circulation, even within neighboring countries, is larger than that described here.

r GAS carriage; 43 different emm types were identified in 82 colonizing GAS isolates (34). Less than one third of emm types identified in our study were also identified in the Ethiopia study, suggesting that the pool of GAS emm types in circulation, even within neighboring countries, is larger than that described here. Reducing the incidence of invasive GAS infection in this setting could be achieved by reducing risk factors such as severe acute malnutrition and HIV (e.g., through prevention of mother-to-child transmission), as well as by supporting antisepsis measures at delivery, including antiseptic neonatal cord care (35–37). Early and improved treatment of skin infections, including impetigo, and burns could also reduce invasive GAS disease. However, prevention through effective vaccination will probably lower disease incidence the most, as has occurred for other pathogens, such as S. pneumoniae (38) and H. influenzae type b (39). The difficulty with emm type–specific GAS vaccine approaches (19) is the heterogeneity of GAS emm types and limited data on many of the emm types identified in this study. From current information, only 57% of invasive GAS disease cases would be covered (either directly or through cross-reactivity) by the most advanced 30-valent vaccine being developed (19). Furthermore, serotype replacement could occur, as described for S. pneumoniae (40), and would require detailed surveillance.

study. From current information, only 57% of invasive GAS disease cases would be covered (either directly or through cross-reactivity) by the most advanced 30-valent vaccine being developed (19). Furthermore, serotype replacement could occur, as described for S. pneumoniae (40), and would require detailed surveillance. The high incidence of invasive GAS disease in rural sub-Saharan Africa underlines the contribution of invasive bacterial disease in this region to childhood deaths, particularly among neonates and young infants; associated case-fatality risk is high. Invasive GAS may also be causing puerperal sepsis in this setting; more studies are needed. Reductions in childhood illness and death could, however, be achieved through effective GAS vaccination. Further development of GAS vaccines followed by clinical trials must be prioritized, targeted at settings with the highest disease incidence. Technical Appendix.  Invasive group A Streptococcus infection among children, Rural Kenya, 1998—2011. Definitions of clinical syndromes, details of Streptococcus pyogenes strains isolated, number of cases, and emm types of isolates. Suggested citation for this article: Seale AC, Davies MR, Anampiu K, Morpeth SC, Nyongesa S, Mwarumba S, et al. Invasive group A Streptococcus infection among children, rural Kenya. Emerg Infect Dis. 2016 Feb [date cited]. http://dx.doi.org/10.3201/eid2202.151358

Technical Appendix.  Invasive group A Streptococcus infection among children, Rural Kenya, 1998—2011. Definitions of clinical syndromes, details of Streptococcus pyogenes strains isolated, number of cases, and emm types of isolates. Suggested citation for this article: Seale AC, Davies MR, Anampiu K, Morpeth SC, Nyongesa S, Mwarumba S, et al. Invasive group A Streptococcus infection among children, rural Kenya. Emerg Infect Dis. 2016 Feb [date cited]. http://dx.doi.org/10.3201/eid2202.151358 Acknowledgments We thank the core informatics, library, and sequencing teams at The Wellcome Sanger Institute for whole-genome sequencing; and we thank all those at the Kenya Medical Research Institute/Wellcome Trust Research Programme who were involved with surveillance at Kilifi County Hospital. We also thank and acknowledge those who set up and managed the MLST global database and the Centers for Disease Control and Prevention global emm database. This study is published with permission from the director of the Kenya Medical Research Institute. This work was supported by the European Society for Paediatric Infectious Disease, The Wellcome Trust, UK (grant nos. 093804, 091758/Z10/Z, 098532, 083579, 077092 to A.C.S., M.R.D., T.N.W., J.A.G.S., G.D., J.B.), and KEMRI-Wellcome Trust and the National Health and Medical Research Council of Australia (grant no. 35250 to M.R.D.). Dr. Seale is a research clinician trained in pediatric infectious diseases and public health. Her research interests are maternal and neonatal infections in sub-Saharan Africa.

During 2014, England and Wales had a sharp increase in the incidence of scarlet fever, which by 2016 had reached 33.2 cases/100,000 person-years, the highest rate in almost 50 years (1,2). An increase in disease incidence was similarly reported from 2009 onward in Vietnam, Singapore, Hong Kong, and mainland China but has not been reported elsewhere in Europe (1,3–7). The cause of this increase is unknown (1,5,8). Scarlet fever was once a common cause of childhood death before incidence and deaths decreased dramatically during the 19th century (1,9). Although now typically a mild disease, scarlet fever remains statutorily reportable in England to enable prediction of periods of increased incidence of invasive group A Streptococcus (iGAS) infection given the temporal correlation between these 2 conditions (1,2). Genomic assessment of Streptococcus pyogenes has furthermore demonstrated that the same strains cause scarlet fever and iGAS infection (10,11). iGAS is statutorily reportable to make contact tracing easier, given the increased risk for secondary iGAS infection among household contacts (4,12,13). This study was initiated as part of a coordinated public health response to determine the cause and effect of the increase in scarlet fever in the United Kingdom (1,2,11). We investigated whether there is an excess risk for secondary iGAS infection in households in which a person was given a diagnosis of scarlet fever to determine whether further public health actions are required to protect contacts.

e the cause and effect of the increase in scarlet fever in the United Kingdom (1,2,11). We investigated whether there is an excess risk for secondary iGAS infection in households in which a person was given a diagnosis of scarlet fever to determine whether further public health actions are required to protect contacts. Methods Study Design, Population, and Definitions We conducted a retrospective cohort study to compare the incidence of iGAS infection among household contacts of persons with scarlet fever with the background incidence of iGAS infection in England. The cohort comprised all scarlet fever case-patients resident in England who had disease onset during January 1, 2011–December 31, 2016. Suspected cases of scarlet fever are reported by clinicians on the basis of clinical signs consistent with the condition, with or without laboratory confirmation of GAS infection. iGAS infection was defined by isolation of GAS from a normally sterile site (including blood, joint aspirates, cerebrospinal/pericardial/peritoneal/pleural fluids, deep tissue or abscess at surgery or necropsy, and bone). A scarlet fever–iGAS household cluster was defined as a household in which a person of any age received a diagnosis of scarlet fever and, within the next 60 days, a different member of the same household received a diagnosis of iGAS infection. Case-patients resident in institutional settings were excluded. A 60-day interval was selected on the basis of preliminary analysis of the interval between onset of scarlet fever and iGAS specimen date in address-matched pairs.

xt 60 days, a different member of the same household received a diagnosis of iGAS infection. Case-patients resident in institutional settings were excluded. A 60-day interval was selected on the basis of preliminary analysis of the interval between onset of scarlet fever and iGAS specimen date in address-matched pairs. Data Sources Demographic details of scarlet fever reports were obtained from the Public Health England (PHE) HP Zone (InFact UK, Ltd., http://hpzoneinfo.in-fact.com), a tool used nationally by health protection teams to assist case and incident management. Reports of iGAS infection were extracted from the PHE national laboratory surveillance database (Second Generation Surveillance System, https://sgss.phe.org.uk). Both datasets were sent to a National Health Service demographic batch tracing service to complete missing postcodes (typically corresponding to 15 addresses [14]), addresses, and patient identifiers. We sought missing postcodes for iGAS cases from the national reference laboratory database. Preliminary analysis indicated that all scarlet fever case-patients within address-matched pairs were <10 years of age. Therefore, data for the number of households in England with >1 child <10 years of age and the number of persons by single year of age in these households were provided by the Office for National Statistics Labour Force Survey (15) for use as denominators in risk calculations. Midyear resident population estimates for 2011 through 2016 were also obtained from the Office for National Statistics.

rs of age and the number of persons by single year of age in these households were provided by the Office for National Statistics Labour Force Survey (15) for use as denominators in risk calculations. Midyear resident population estimates for 2011 through 2016 were also obtained from the Office for National Statistics. We obtained clinical information for cases within scarlet fever–iGAS pairs from HP Zone. National laboratory surveillance and reference laboratory data were used to identify co-infection and emm typing. We obtained the index of multiple deprivation decile score for each household on the basis of residential postcode (16). This index is a measure of relative socioeconomic deprivation based on 7 domains and provides a ranking at granular geographic level (≈1,500 residents) from the most to least deprived areas in England (17). Data Analysis Identification of Household Clusters We cleaned and analyzed data by using R version 3.2.2 (https://cran.r-project.org/bin/windows/base/old/3.2.2). Records without a postcode were excluded. We matched scarlet fever cases to all iGAS reports with a specimen date during November 1, 2010–March 1, 2017, by residential postcode. This matching enabled capture of linked iGAS cases occurring within 2 months of the first and last scarlet fever case. Cases within postcode-matched pairs without full address were removed. Addresses of remaining matched pairs with an interval between onset of scarlet fever and iGAS specimen date <60 days were visually scrutinized to exclude institutional settings and confirm co-location.

onths of the first and last scarlet fever case. Cases within postcode-matched pairs without full address were removed. Addresses of remaining matched pairs with an interval between onset of scarlet fever and iGAS specimen date <60 days were visually scrutinized to exclude institutional settings and confirm co-location. The iGAS and scarlet fever datasets were deduplicated after matching to ensure that all sequential specimens were considered in identifying temporal links between cases. An interval >14 days between specimen dates was considered a new episode for iGAS and >30 days between onset dates for scarlet fever. To supplement household clusters identified from address matching, we reviewed GAS clusters and outbreaks recorded on HP Zone and the reference laboratory outbreak database over the same period.

val >14 days between specimen dates was considered a new episode for iGAS and >30 days between onset dates for scarlet fever. To supplement household clusters identified from address matching, we reviewed GAS clusters and outbreaks recorded on HP Zone and the reference laboratory outbreak database over the same period. Calculation of Risk We estimated the average number of household contacts of scarlet fever cases by dividing the total number of persons living in households with a child <10 years of age in England during the study period by the number of households and subtracting 1 to account for the case-patient being a household member. We multiplied this figure by the number of scarlet fever cases to estimate the total number of contacts and calculated the person-years at risk over 60 days. We calculated the incidence of iGAS infection among scarlet fever household contacts by dividing the number of iGAS cases among these contacts by the number of person-years at risk and used Poisson distribution to define 95% CIs. The background rate of iGAS infection was based on the total number of iGAS cases in England. We repeated this analysis by year and age group. We conducted a sensitivity analysis to investigate the effect of increasing the average household size by up to 3 household members (3.8/household to 6.8/household).

ne 95% CIs. The background rate of iGAS infection was based on the total number of iGAS cases in England. We repeated this analysis by year and age group. We conducted a sensitivity analysis to investigate the effect of increasing the average household size by up to 3 household members (3.8/household to 6.8/household). Ethics Considerations Ethics approval was not required because we analyzed only routinely collected data. PHE has the authority to collect and process confidential patient information for communicable disease surveillance and control under Section 251 of the National Health Service Act of 2006. Results A total of 73,344 scarlet fever cases were reported to PHE during 2011–2016. Of the 9,978 episodes of iGAS infection extracted for address matching to scarlet fever cases, 2.7% (269) were excluded due to a missing address; a higher proportion of cases before 2014 had a missing postcode than cases from 2014 onward (4.4% vs. 1.4%, χ2 85.2, df 1, 95% CI 2.3–3.7; p<0.0001). We identified 991 scarlet fever–iGAS pairs with identical postcodes in any setting (including institutions); 1.8% (18) did not have a full address and were excluded from further analysis (onset interval range 93–1,893 days) (Figure 1). Of the remaining 973 pairs, 53 were resident in a private home and confirmed as being at the same address; iGAS cases occurred after scarlet fever onset for 28 of 53 pairs.

uding institutions); 1.8% (18) did not have a full address and were excluded from further analysis (onset interval range 93–1,893 days) (Figure 1). Of the remaining 973 pairs, 53 were resident in a private home and confirmed as being at the same address; iGAS cases occurred after scarlet fever onset for 28 of 53 pairs. Figure 1 Summary of records included at each stage of the matching process of scarlet fever and iGAS cases, England 2011–2016. *Interval between excluded pairs was >60 days. †A household cluster was defined on the basis of a person being given a diagnosis of scarlet fever and a different member of the same household given a diagnosis of iGAS infection for which onset of iGAS symptoms occurred within 60 days after onset of scarlet fever. iGAS, invasive group A Streptococcus infection.

days. †A household cluster was defined on the basis of a person being given a diagnosis of scarlet fever and a different member of the same household given a diagnosis of iGAS infection for which onset of iGAS symptoms occurred within 60 days after onset of scarlet fever. iGAS, invasive group A Streptococcus infection. A pronounced increase in the number of pairs was evident within the first 100 days after onset of scarlet fever (Figure 2): 13 pairs identified, compared with an expected 1.5 (95% CI 0.2–7.2) iGAS cases based on background iGAS infection rates. All 13 pairs were within 60 days, and on review of case details, 11 met the household cluster definition. No clusters were identified through review of the national case management system or the reference laboratory database. Two of the 25 pairs in which iGAS occurred before scarlet fever had an interval between cases <60 days, a rate of 6.4 iGAS cases/100,000 person-years and twice the background rate (rate ratio [RR] 2.2, 95% CI 0.6–8.9). Fifty-one pairs with onset dates within 60 days had the same postcode but were resident in different private homes.

pairs in which iGAS occurred before scarlet fever had an interval between cases <60 days, a rate of 6.4 iGAS cases/100,000 person-years and twice the background rate (rate ratio [RR] 2.2, 95% CI 0.6–8.9). Fifty-one pairs with onset dates within 60 days had the same postcode but were resident in different private homes. Figure 2 Distribution of time interval between onset of scarlet fever and iGAS within address-matched pairs (n = 53) and expected number of clusters, England 2011–2016. Exploratory analysis was used to identify the period of excess numbers of iGAS cases before review of case records; iGAS cases might be linked to >1 scarlet fever case episode in the same household. The background iGAS rate was 2.88 cases/100,000 person-years; 95% CI is based on 2 expected cases/100 days. There were 189,684 scarlet fever household contacts. iGAS, invasive group A Streptococcus infection. We identified 18 persons given a diagnosis of iGAS and scarlet fever. The median interval between onset of scarlet fever and iGAS specimen date was 155 days (range 5–1,488 days). Sixteen cases had an interval >100 days, and 7 cases had iGAS infection after a diagnosis of scarlet fever.

Figure 2 Distribution of time interval between onset of scarlet fever and iGAS within address-matched pairs (n = 53) and expected number of clusters, England 2011–2016. Exploratory analysis was used to identify the period of excess numbers of iGAS cases before review of case records; iGAS cases might be linked to >1 scarlet fever case episode in the same household. The background iGAS rate was 2.88 cases/100,000 person-years; 95% CI is based on 2 expected cases/100 days. There were 189,684 scarlet fever household contacts. iGAS, invasive group A Streptococcus infection. We identified 18 persons given a diagnosis of iGAS and scarlet fever. The median interval between onset of scarlet fever and iGAS specimen date was 155 days (range 5–1,488 days). Sixteen cases had an interval >100 days, and 7 cases had iGAS infection after a diagnosis of scarlet fever. Characteristics of Household Clusters All household clusters were composed of 1 scarlet fever case and 1 iGAS case and occurred after March 2014. The median interval between onset of scarlet fever and of iGAS infection was 18 days (range 3–54 days) (Figure 3). Five iGAS cases occurred in parents of children with scarlet fever and 4 in siblings; the relationship to the scarlet fever case-patient was not recorded for 2 clusters (iGAS case-patients 86 and 26 years of age) (Table 1). Five iGAS case-patients had sepsis and 3 had cellulitis. Five of the 11 iGAS case-patients had an underlying chronic condition, and 1 had an acute infection with influenza. Three of the iGAS case-patients died; 2 of them had predisposing conditions (arthritis, diabetes, and atypical mycobacterial infection). The median household size was 4 persons (range 4–6 persons). Seven of 11 households were in the 30% most deprived neighborhoods in England, and 3 were in the 30% least deprived.

enza. Three of the iGAS case-patients died; 2 of them had predisposing conditions (arthritis, diabetes, and atypical mycobacterial infection). The median household size was 4 persons (range 4–6 persons). Seven of 11 households were in the 30% most deprived neighborhoods in England, and 3 were in the 30% least deprived. Figure 3 Distribution of time intervals between onset of scarlet fever and invasive group A Streptococcus infection within each pair meeting the household cluster definition (n = 11), England 2011–2016. Table 1 Characteristics of 11 iGAS case-patients within household clusters, England, 2011–2016* Characteristic           No. case-patients Total Male sex Female sex Age, y <1 2 1 1 1–18 2 0 2 19–50 6 3 3 >75 1 1 0 Total 11 5 6 Relationship to scarlet fever case-patient Parent 5 3 2 Sibling 4 1 3 Unknown 2 1 1 Acute health conditions at time of diagnosis Influenza A 1 NR NR Chronic health condition at time of diagnosis Arthritis 1 NR NR Crohn’s disease 1 NR NR Premature birth 1 NR NR Diabetes 1 NR NR Atypical mycobacterial infection 1 NR NR Asplenia 1 NR NR Multiple unnamed conditions 1 NR NR No concurrent conditions 6 NR NR Died 3 NR NR Clinical manifestation Sepsis 5 NR NR Cellulitis 3 NR NR Septic arthritis 1 NR NR Other invasive infection (unspecified) 2 NR NR iGAS emm typing 1.0 4 NR NR 4.0 1 NR NR 12.0 1 NR NR Untyped 6 NR NR *iGAS, invasive group A Streptococcus infection; NR, not reported.

s 1 NR NR No concurrent conditions 6 NR NR Died 3 NR NR Clinical manifestation Sepsis 5 NR NR Cellulitis 3 NR NR Septic arthritis 1 NR NR Other invasive infection (unspecified) 2 NR NR iGAS emm typing 1.0 4 NR NR 4.0 1 NR NR 12.0 1 NR NR Untyped 6 NR NR *iGAS, invasive group A Streptococcus infection; NR, not reported. Strain typing was available for 6 of the iGAS household cluster cases, 4 of which were emm 1.0 and 1 each were emm 4.0 and emm 12.0. Typing results for scarlet fever isolates were not available.

s 1 NR NR No concurrent conditions 6 NR NR Died 3 NR NR Clinical manifestation Sepsis 5 NR NR Cellulitis 3 NR NR Septic arthritis 1 NR NR Other invasive infection (unspecified) 2 NR NR iGAS emm typing 1.0 4 NR NR 4.0 1 NR NR 12.0 1 NR NR Untyped 6 NR NR *iGAS, invasive group A Streptococcus infection; NR, not reported. Strain typing was available for 6 of the iGAS household cluster cases, 4 of which were emm 1.0 and 1 each were emm 4.0 and emm 12.0. Typing results for scarlet fever isolates were not available. Calculation of Risk All scarlet fever case-patients within clusters were <10 years of age. Therefore, we restricted analysis of risk for iGAS infection to household contacts of scarlet fever case-patients <10 years of age (n = 66,191). We estimated that these case-patients had 189,684 household contacts (average household size 3.9 persons). We estimated the incidence of iGAS infection among these contacts to be 35.3 cases/100,000 person-years (95% CI 17.6–63.2 cases/100,000 person-years) compared with a background incidence of iGAS in England (all ages) of 2.9 cases/100,000 person-years (Table 2). Therefore, the rate of iGAS infection in household contacts of persons with scarlet fever was 12 times higher than the background rate in England over the same period (RR 12.2, 95% CI 6.7–22.1) (Table 2). The highest absolute rates were for infants (138 cases/100,000 person-years, 95% CI 16.7–496.8 cases/100,000 person-years) and persons >75 years of age (1,419 cases/100,000 person-years, 95% CI 35.9–7907.3 cases/100,000 person-years), although these rates were based on a small number of cases (Table 3). RR was highest for contacts 11–17 years of age (RR 43.9, 95% CI 6.1–313.7) and contacts >75 years of age (RR 139.2, 95% CI 19.6–988.5) (Table 3).

sons >75 years of age (1,419 cases/100,000 person-years, 95% CI 35.9–7907.3 cases/100,000 person-years), although these rates were based on a small number of cases (Table 3). RR was highest for contacts 11–17 years of age (RR 43.9, 95% CI 6.1–313.7) and contacts >75 years of age (RR 139.2, 95% CI 19.6–988.5) (Table 3). Table 2 Risk for iGAS infection among household contacts of scarlet fever case-patients <10 years of age by year compared with background iGAS incidence, England, 2011–2016* Year No. scarlet fever cases Estimated no. contacts No. iGAS cases in contacts† Attack rate/100,000 person-years (95% CI) Background iGAS incidence/100,000 person-years Rate ratio (95% CI) 2011 3,128 8,929 0 0.0 (0.0–251.5) 2.29 NA 2012 4,632 13,073 0 0.0 (0.0–171.8) 2.35 NA 2013 5,204 14,778 0 0.0 (0.0–152.0) 2.99 NA 2014 16,394 47,015 3 38.8 (8.0–113.5) 2.29 16.9 (5.5–52.6) 2015 18,022 51,454 3 35.5 (7.3–103.7) 3.48 10.2 (3.3–31.7) 2016 18,811 54,435 5 55.9 (18.2–130.5) 3.91 14.3 (5.9–34.7) Total 66,191 189,684 11 35.3 (17.6–63.2) 2.89 12.2 (6.7–22.1) *iGAS, invasive group A Streptococcus infection; NA, not applicable. †During the 60 days after onset of scarlet fever in the household.

.6) 2015 18,022 51,454 3 35.5 (7.3–103.7) 3.48 10.2 (3.3–31.7) 2016 18,811 54,435 5 55.9 (18.2–130.5) 3.91 14.3 (5.9–34.7) Total 66,191 189,684 11 35.3 (17.6–63.2) 2.89 12.2 (6.7–22.1) *iGAS, invasive group A Streptococcus infection; NA, not applicable. †During the 60 days after onset of scarlet fever in the household. Table 3 Risk for iGAS among household contacts of 66,191 scarlet fever case-patients <10 years of age compared with background iGAS incidence by age, England, 2011–2016* Age of contacts, y Estimated no. contacts No. iGAS cases in contacts† Attack rate/100,000 person-years (95% CI) Background iGAS incidence/100,000 person-years Rate ratio (95% CI) <1 8,853 2 137.5 (16.7–496.8) 6.42 21.4 (5.31–86.1) 1–10 28,660 1 21.2 (0.5–118.3) 2.84 7.5 (1.1–53.1) 11–17 22,209 1 27.4 (0.7–152.7) 0.58 43.9 (6.1–313.7) 18–50 122,801 6 29.7 (10.9–64.7) 1.69 18.4 (8.4–41.1) 51–74 6,733 0 0 (0–333.5) 3.59 0 >75 429 1 1,419.2 (35.9–7.907.3) 10.20 139.2 (19.6–988.5) iGAS, invasive group A Streptococcus infection. †During the 60 days after onset of scarlet fever in the household. Sensitivity analysis, increasing the average household size from 3.8 to 6.8 members, reduced the RR for iGAS in scarlet fever household contacts relative to the background iGAS rate to 6 (95% CI 3.3–10.8). The rate of iGAS infection in scarlet fever household contacts before the period of increased scarlet fever incidence (2011–2013), 0 cases/100,000 person-years (95% CI 0.0–61.1), was not significantly different for the period of increased incidence (2014–2016), 43.7 cases/100,000 person-years (95% CI 21.8–78.4).

10.8). The rate of iGAS infection in scarlet fever household contacts before the period of increased scarlet fever incidence (2011–2013), 0 cases/100,000 person-years (95% CI 0.0–61.1), was not significantly different for the period of increased incidence (2014–2016), 43.7 cases/100,000 person-years (95% CI 21.8–78.4). Discussion Our study identified a low risk for iGAS infection among household contacts of scarlet fever cases (35.3 cases/100,000 person-years). However, this risk was increased when compared with the background risk. Eleven iGAS cases occurred among ≈189,684 contacts during the 60 days after scarlet fever onset; 1.5 cases would have been expected on the basis of a background rate of 2.9 cases/100,000 person-years. Small numbers of cases preclude robust subgroup analysis but excess risk was highest in contacts >75 and 11–17 years of age. Although the absolute risk was low, the effect of these infections was severe; 3 deaths were reported. These findings have implications for other countries reporting a high incidence of scarlet fever.

numbers of cases preclude robust subgroup analysis but excess risk was highest in contacts >75 and 11–17 years of age. Although the absolute risk was low, the effect of these infections was severe; 3 deaths were reported. These findings have implications for other countries reporting a high incidence of scarlet fever. Half the secondary iGAS cases occurred in parents and one third occurred in siblings of scarlet fever case-patients. A previous study showed a slight excess in scarlet fever incidence in young adult women, possibly explained by caring responsibilities for children with the infection (1). We did not observe a similar pattern for iGAS; 3 of 5 cases in parents were in men. Contacts >75 years of age had the highest absolute risk for development of iGAS. Although the background rate for iGAS was highest in elderly persons, because there were only an estimated 429 household contacts of scarlet fever case-patients within this age group, 1 secondary iGAS case in this group translated into a high attack rate.

rs of age had the highest absolute risk for development of iGAS. Although the background rate for iGAS was highest in elderly persons, because there were only an estimated 429 household contacts of scarlet fever case-patients within this age group, 1 secondary iGAS case in this group translated into a high attack rate. An estimated 5 million grandparents have regular childcare responsibilities in the United Kingdom (18). We were unable to assess the risk for grandparents not living in the same household and are likely missing a major group potentially at risk through contact with scarlet fever case-patients. Our identification of 51 postcode-matched pairs with different addresses suggests a possible increased risk for iGAS infection in the neighborhood of scarlet fever case-patients and warrants further assessment. Although a proportion of the observed secondary household iGAS risk might be caused by transmission in wider social networks, the fact that parents constituted most secondary iGAS cases suggests that transmission within the home underpins these clusters because parents are less likely than children to be exposed to scarlet fever outside the home.

observed secondary household iGAS risk might be caused by transmission in wider social networks, the fact that parents constituted most secondary iGAS cases suggests that transmission within the home underpins these clusters because parents are less likely than children to be exposed to scarlet fever outside the home. Almost half the iGAS case-patients reported underlying chronic conditions, although diabetes, Crohn’s disease, and arthritis are common, which limits the potential to target public health actions. A broad range of clinical initial manifestations were reported for iGAS infections within clusters, including skin and soft tissue and joint infections. Overcrowding is a known predisposing factor for S. pyogenes infections (4), but we did not find evidence of this factor in clusters who lived in average-sized households (median 4 occupants), although this information was not available in public health records for 4 of 11 households. Pharyngeal carriage of GAS among close contacts of persons with invasive infections has been demonstrated (19); person-to-person transmission of GAS occurs by respiratory droplets or skin contact (4). We do not assume that iGAS infection was necessarily acquired from the child with scarlet fever. Other members of the household (symptomatic or asymptomatic) might have been the source to either case-patient, particularly given the long period of risk (60 days) and that the scarlet fever case-patient probably received treatment.

assume that iGAS infection was necessarily acquired from the child with scarlet fever. Other members of the household (symptomatic or asymptomatic) might have been the source to either case-patient, particularly given the long period of risk (60 days) and that the scarlet fever case-patient probably received treatment. We examined the risk for secondary iGAS infection before and after onset of scarlet fever, without preconceptions as to the length of identified period of excess risk. Transmission of GAS within the household for 60 days is plausible; back-and-forth transmission between household members is well-described (20–22) and has been demonstrated to occur over a 10-month period (21). Transmission from an asymptomatic carrier can occur up to several weeks after acquisition although communicability is lower than from symptomatic cases (19,23). An ongoing study in London aims to assess GAS carriage in family members of scarlet fever case-patients (24). Environmental reservoirs have been implicated in hospital (25,26), nursery (27), and care home outbreaks (4,28), but the duration of viability in the environment is unknown. Survival on dry surfaces has been demonstrated after several months; therefore, public health messaging should include advice on infection control in households, particularly where there are susceptible persons (29,30).

y (27), and care home outbreaks (4,28), but the duration of viability in the environment is unknown. Survival on dry surfaces has been demonstrated after several months; therefore, public health messaging should include advice on infection control in households, particularly where there are susceptible persons (29,30). Although it was not the focus of this study, we observed a slight excess risk for iGAS occurring before scarlet fever. Public health guidelines on the management of iGAS infection in the United Kingdom recommend that household contacts are advised to visit their general practitioner (GP) for assessment if they have symptoms of GAS infection in the 30 days after onset in the index case-patient. Therefore, these scarlet fever case-patients should have already been under surveillance, potentially increasing the likelihood of their diagnosis. The number of contacts of scarlet fever case-patients was the main source of uncertainty in our risk estimation. If households that have scarlet fever case-patients differ in size from the average household with children, we could have underestimated or overestimated the risk. However, the risk for iGAS infection would still have been 6 times higher than background if the average household size was increased to 7 members. Coupled with our observation that household clusters had a median size of 4 (compared with 3.9 for all households in England), it is unlikely the uncertainty about numbers of contacts could account for the increase in risk observed.

times higher than background if the average household size was increased to 7 members. Coupled with our observation that household clusters had a median size of 4 (compared with 3.9 for all households in England), it is unlikely the uncertainty about numbers of contacts could account for the increase in risk observed. Failure to match iGAS to scarlet fever cases could have occurred because of missing postcodes (3% of iGAS cases), errors in the postcode or address, or because the traced postcode represents the current address and might be different from that at the time of infection. Our finding that all clusters occurred after the increase in scarlet fever during 2014 was possibly influenced by this factor, given that postcode completion was higher in the later years of the study and enabled identification of clusters. We did not adjust for residence in a long-term care facility or hospitalization in the background risk calculation: 3.5% of iGAS cases in England (2009–2010) were estimated to be acquired in long-term care facilities, and 6% of these infections were estimated to be acquired in hospitals. Residents of long-term care facilities had a 6-fold higher risk for iGAS infection than community residents (31,32). Including institutionally acquired infections slightly increased the background iGAS risk in this study.

long-term care facilities, and 6% of these infections were estimated to be acquired in hospitals. Residents of long-term care facilities had a 6-fold higher risk for iGAS infection than community residents (31,32). Including institutionally acquired infections slightly increased the background iGAS risk in this study. We used clinical reports of scarlet fever and recognized that a proportion of reported cases might have had other infections, which has the potential to influence the risk estimate in either direction. Although only ≈50% of scarlet fever consultations in primary care in England are formally reported (1), this finding would not influence the risk estimate because it was based on iGAS cases that occurred in contacts of only the reported cohort. We did not capture the burden of disease associated with severe noninvasive GAS infections (GAS isolated from a nonsterile site), although these infections are estimated to comprise only 1% of all severe GAS infections (33,34).

te because it was based on iGAS cases that occurred in contacts of only the reported cohort. We did not capture the burden of disease associated with severe noninvasive GAS infections (GAS isolated from a nonsterile site), although these infections are estimated to comprise only 1% of all severe GAS infections (33,34). We assessed the risk for secondary iGAS infection for household contacts to be low. However, potential severity is high. The prodrome for iGAS infection can be nonspecific, and the disease can progress rapidly; therefore, increasing the index of suspicion in specific circumstances or groups at increased risk might expedite diagnosis and commencement of life-saving treatment (4,35). Offering antimicrobial drug prophylaxis to household contacts to eradicate carriage and treat incipient infection could reduce the risk for iGAS infection. However, the unintended consequences of large-scale increased use of antimicrobial drugs, heightened patient anxiety, the effect on GP workload, and the lack of evidence on effectiveness make this option disproportionate given the low overall risk estimated. Antimicrobial drug prophylaxis could be targeted to high-risk contacts, such as elderly persons and infants. However, there is considerable uncertainty for the risk estimate for these groups because of the small number of secondary iGAS cases. Providing information on signs and symptoms of iGAS infection to patients or parents at the point of scarlet fever diagnosis to accelerate self-referral for medical assessment could be effective but has the potential to increase anxiety for many persons and increase presentations of worried healthy persons to GPs and emergency departments at scale. Increasing awareness among frontline clinicians assessing patients of this increased risk to improve early identification and treatment of cases is perhaps the most proportionate and feasible response on the basis of available data. Information could also be made available to the public through patient-facing websites provided that messages are worded carefully so as not to increase anxiety.

increased risk to improve early identification and treatment of cases is perhaps the most proportionate and feasible response on the basis of available data. Information could also be made available to the public through patient-facing websites provided that messages are worded carefully so as not to increase anxiety. We recommend repeating this analysis at regular intervals to monitor and increase precision around our estimated risk. Enhanced surveillance of iGAS patients should include questions on the number of contacts and recent scarlet fever infections in the household. This information would help address some of the methodological uncertainties around the number of contacts and enable assessment of the attributable risk in the context of other risk factors. Of ≈10,000 iGAS cases identified during our study, only 11 were associated with scarlet fever contact: as such, a proportionate response to further investigations is warranted. Although increases in iGAS infection have been observed during the latter period of the scarlet fever upsurge (2016 onward), these increases follow a longer-term trend of increasing iGAS infection in England, and the connection with increased scarlet fever activity remains unclear (36).

her investigations is warranted. Although increases in iGAS infection have been observed during the latter period of the scarlet fever upsurge (2016 onward), these increases follow a longer-term trend of increasing iGAS infection in England, and the connection with increased scarlet fever activity remains unclear (36). It is likely that contact with other superficial manifestation of GAS infection would also increase the risk for iGAS infection. However, the mixed etiology for these conditions and lack of microbiological testing make this potential risk difficult to assess. Nonetheless, with drives to reduce antimicrobial drug treatment for conditions such as pharyngitis to relieve selection pressure favoring antimicrobial drug resistance, understanding the possible repercussions for the patient and wider community are essential. In conclusion, we identified an excess risk for iGAS among household contacts of scarlet fever case-patients, although we assessed the overall risk to be low. We recommend that frontline clinicians maintain heightened awareness of the risk for iGAS in scarlet fever contacts when assessing patients. Further research to tighten our risk estimates and improve our understanding of transmission patterns in households will inform future prevention strategies.

erall risk to be low. We recommend that frontline clinicians maintain heightened awareness of the risk for iGAS in scarlet fever contacts when assessing patients. Further research to tighten our risk estimates and improve our understanding of transmission patterns in households will inform future prevention strategies. Suggested citation for this article: Watts V, Balasegaram S, Brown CS, Mathew S, Mearkle R, Ready D, et al. Increased risk for invasive group A Streptococcus disease for household contacts of scarlet fever cases, England, 2011–2016. Emerg Infect Dis. 2019 Mar [date cited]. https://doi.org/10.3201/eid2503181518 Acknowledgments We thank the National Scarlet Fever Incident Management Team and the Demographic Analysis Unit at the Office for National Statistics for supplying population data; Cliff Lake for arranging access to the national case management database; Nick Hinton for extracting iGAS case data; Health Protection Teams for expert management of cases and outbreaks; microbiology laboratories for submission of isolates for testing; and clinicians and other infectious disease reporters for providing information. Ms. Watts is a field epidemiology training fellow with the Field Service, Public Health England, Liverpool, UK. Her research interests include surveillance and epidemiology of S. pyogenes infections.